Breaking Free From Fusion Rule: A Fully Semantic-Driven Infrared and Visible Image Fusion

Infrared and visible image fusion plays a vital role in the field of computer vision. Previous approaches make efforts to design various rules loss functions. However, these experimental designed methods more complex. Besides, most them only focus on boosting visual effects, thus showing unsatisfactory performance for follow-up high-level vision tasks. To address challenges, this letter, we develop semantic-level network sufficiently utilize semantic guidance, emancipating rules. In addition, achieve better understanding feature process, block based transformer is presented multi-scale manner. Moreover, devise regularization function, together with training strategy, fully use guidance from Compared state-of-the-art methods, our method does not depend hand-crafted function. Still, it achieves superior quality along